At Rise of AI 2023 in Berlin, Sabina Jeschke, CEO of KI Park e.V. and CIO at Quantagonia, argues that quantum computing is the next technology, alongside 5G/6G connectivity, that will push artificial intelligence beyond today's algorithm-driven phase of deep learning and reinforcement learning. She debunks four common assumptions: that quantum computing is too far away (deep-temperature systems reach industrial scale by 2025, room-temperature systems by 2028), that it only applies to niche research, that no usable quantum computers exist yet, and that it is too expensive. She frames quantum computing as unlocking problems that classical computing cannot execute in time, structured around optimization, simulation and AI, and illustrates this with cases from liquid gas routing, aircraft maintenance logistics, finance, fraud detection and reinforcement learning for autonomous driving. She closes by linking quantum computing to sustainability through orbital solar power, quantum-optimized flight paths and lower long-term energy use than commonly assumed.
Key takeaways
- Deep-temperature quantum computing is expected to reach industrial scale by 2025, with room-temperature, cooling-free systems following by 2028.
- A full train-schedule simulation across tracks and stations takes two to three weeks to compute, so it supports planning but not today's real-time rescheduling.
- IBM and ExxonMobil used quantum annealers to optimize liquid gas shipping and storage routing.
- Quantum-accelerated reinforcement learning can speed up image recognition enough to make certain autonomous-driving features usable in production.
- Industry estimates put quantum computing's energy use at roughly 10 percent of today's systems, and 2-5 percent for room-temperature systems.
Detailed notes
Quantum computing as the next leap for AI
Jeschke frames the present as a fourth industrial revolution driven by artificial intelligence, currently in an "algorithm-driven phase" of reinforcement learning, deep learning, Q-learning and supervised learning. She argues two further technologies will push AI to its next level: 5G/6G connectivity, whose low latency lets autonomous systems shift computation from the vehicle into the cloud within about six years, and high-performance quantum computing, which is the subject of her talk. She structures the talk around debunking wrong assumptions about quantum computing, explaining from a "helicopter perspective" what is actually changing, giving real-world examples, and finally connecting quantum computing to sustainability.
Four wrong assumptions about quantum computing
First, that quantum computing is too far away: Jeschke cites projections that deep-temperature quantum computing stacks will break through at industrial scale by 2025 and room-temperature systems that need no cooling by 2028, compared with the years German industry took to adopt AI or blockchain. Second, that it only matters for niche research such as protein folding or climate modeling: she gives a rail scheduling example where a full simulation of tracks, stations and trains takes two to three weeks of compute, useful for analytics but not for the real-time rescheduling needed when disruptions occur three to four times a minute. Third, that no usable quantum computers exist today: organizations can already prepare by hiring the right people, building ecosystems, and using annealers or quantum-inspired algorithms written for eventual quantum hardware. Fourth, that it is too expensive: she expects early deep-temperature quantum computing to be offered as a cloud license rather than owned hardware, making it affordable.
From mainframes to PCs to quantum: expanding access
Jeschke compares the shift to quantum computing with earlier hardware transitions that mattered less for raw computing power than for who could access it: moving from centralized computer centers to the PC opened computing to a broad population, and mobile internet then made computing available anywhere rather than only in fixed locations. She argues quantum computing represents a similar expansion of what can be computed at all, though not everything and not in its first years. She also notes that Moore's Law is approaching its end, possibly by 2025 or 2030, due to the limits of miniaturization, which she frames as one reason quantum computing's timing matters.
Four capability leaps and business use cases
Jeschke lists four capability shifts she expects from quantum computing: executing simulations too large to run in practical time on classical computers; steering systems in real time rather than only planning schedules offline, as in her train example; finding truly optimal rather than merely acceptable solutions within a fixed computation budget; and running physically accurate digital twins at scale to speed up product design, particularly in chemistry and drug design. She illustrates these with cases she has personally seen: IBM and ExxonMobil using quantum annealers to optimize liquid gas shipping and storage routing; aircraft maintenance logistics solved with a D-Wave annealer as a variant of the traveling salesman problem; early use of quantum methods in finance and high-frequency trading; and fraud and cyberattack detection modeled like an immune system identifying and isolating intrusions. A former Volkswagen colleague, now at a quantum computing company, showed her that quantum-accelerated reinforcement learning can speed up image recognition tasks, such as recognizing a familiar route under different weather conditions, enough to make them usable in production for autonomous driving.
Sustainability payoff
Jeschke argues quantum computing will contribute to sustainability in three ways: faster innovation cycles, optimization that reduces climate impact alongside cost, and lower energy consumption than commonly assumed. She cites a European Space Agency project on orbital solar power plants, estimating that about 25 such installations could cover Europe's energy consumption, a small fraction compared with more than 100 nuclear power plants, where quantum computing could compress decades of engineering into a few years. She also describes a Deloitte AI Pact challenge in which startups used quantum computing to recompute aircraft flight paths in three dimensions every few seconds to reduce or reverse the warming effect of condensation trails, a calculation she says is infeasible with today's hardware. On energy use, she states that despite the need to cool deep-temperature systems, industry estimates put quantum computing at around 10 percent of today's energy consumption, falling to 2-5 percent for room-temperature systems, though she stresses these figures are estimates from industry contacts rather than a published study.
Q&A highlights
Asked what could be quantum computing's "ChatGPT moment," Jeschke says large language models may already be lowering the barrier to programming quantum computers: she describes asking ChatGPT to write code for a traveling salesman problem and then in Qiskit, suggesting the programming-language barrier is being addressed while cloud providers such as AWS are starting to offer access to real, if still limited, quantum hardware. Asked whether quantum computing could also help bring AI itself to a higher level, she points to reinforcement learning as a way for European industry to avoid dependence on large datasets. Joanna Bryson raises two concerns: the safety of real-time autonomous vehicle control over 5G if the connection is disrupted or attacked, to which Jeschke responds that cars need a fast onboard "little brain" for immediate actions like braking plus a cloud-based "big brain" for the rest, with fallback logic if the connection fails; and the actual empirical power usage of quantum computing, to which Jeschke says she does not yet have a formal study but is working on one, and offers the same estimated figures discussed earlier as the best current approximation.
Quotes
Whenever a new technology breaks through, it's kind of expensive.
Sabina Jeschke
I would like to switch my iPhone to a Q-phone, not because I need the computational power, but because of sustainability reasons, because the energy consumption is so much lower.
Sabina Jeschke
There are two barriers: the first is where's the hardware, and the second is that you cannot program them in traditional languages.
Sabina Jeschke
Questions answered in this talk
When will quantum computing reach industrial scale?
According to Sabina Jeschke at Rise of AI 2023, deep-temperature quantum computing will break through at industrial scale by 2025, followed by room-temperature systems that need no cooling by 2028.
Does quantum computing use more energy than traditional computing?
According to Sabina Jeschke at Rise of AI 2023, despite the need to cool deep-temperature systems, industry estimates suggest quantum computers will use only about 10 percent of the energy of today's systems, dropping to 2-5 percent for room-temperature systems, though she notes these figures are not yet from a published study.
How can quantum computing make autonomous driving safer or more capable?
According to Sabina Jeschke at Rise of AI 2023, image recognition tasks that are currently too slow for production use, such as recognizing a familiar route in changing weather, can be sped up with quantum-accelerated reinforcement learning, and real-time vehicle control can combine a fast onboard system with a cloud-based system with fallback logic.